Many agents, one console.
Each agent runs OpenClaw on its own server. The console shows all of them: what each one is doing, how much AI they use, and what is waiting for you. Start new agents from a template, and give them the apps they need.
Start from a template.
A template sets up a new agent for a job: its instructions, skills and settings. Skills are shared, so you write them once for all your agents. With GBrain, your agents share one memory of past decisions and client details. Export an agent's workspace any time, as one tarball.
Templates to start from
Pick Personal assistant, Software factory, Chief of staff, Support agent, or Blank. Save an agent you set up as a template of your own. Create several agents at once from one template, for example one for each person on a team.
Skills you write once
A skill is a set of steps an agent follows for one task. Write one, upload one, or bring one in from ClawHub or GitHub. Every change is a new version, and you choose which agents have it.
Your apps, without handing over keys.
Connect an app once for your organization, then choose which agents may use it. You sign in once. The firewall keeps the login, and the agents get a stand-in. You can also manage ControlClaw itself from Claude or another MCP client.
About 1,500 apps
Connect HubSpot, Notion, Xero, GitHub and about 1,500 more apps through Open Connector, which runs on your firewall. Or add any remote MCP server, the standard way AI tools connect to other apps. Remove an app and every agent loses it at once.
Rules for what agents do in apps
Block an action, like deleting a contact, or ask a person before it runs. Or write the rule as a sentence, like "Ask before sharing a folder outside the company", and the firewall's AI checks each call. Every new organization starts with one rule. It asks before anything that deletes.
